For a manufacturer managing complex technical products, multilingual content, frequent acquisitions, and millions of possible product configurations, giving every team fast access to reliable product information is not a small data project.

DwyerOmega is using Akeneo’s Model Context Protocol (MCP) to connect product information from its Akeneo PIM to Microsoft Fabric, Power BI, AI tools, SharePoint, and AWS. The result is a growing operational layer that makes product data easier to access, analyze, improve, and activate.

Instead of asking every product manager or marketer to learn another system, DwyerOmega brings governed product information into the tools teams already use. Behind the scenes, MCP also helps the company replace development-heavy integrations with more flexible workflows for data extraction, content enrichment, quality monitoring, and acquisition onboarding.

The challenge: trusted data was available, but not always accessible

DwyerOmega had already established the Akeneo PIM as the source of truth for product information. The next challenge was making that information useful to teams with different responsibilities, habits, and technical skills.

Cyrus Grimes, Director of Product Content Management at DwyerOmega, had been extracting enrichment data through API connections to build Power BI dashboards. The approach worked, but it carried a cost.

Product managers run business units and balance many priorities. Marketing and SEO teams need to know where descriptions, metadata, images, and supporting content are missing. Sales operations need visibility into available catalog SKUs. These stakeholders benefit from Akeneo data, but they do not all work in the PIM every day.

“The very concept of opening another software program to access information, I felt, was a blocker.”

The pipeline itself was also reaching its limit. Multilingual content compounded the volume of every full load, and the extraction into Microsoft Fabric was not stable enough to support the next stage of growth.

The solution: an MCP-powered product data pipeline

Using Akeneo MCP with Codex inside Microsoft Fabric, Cyrus rebuilt the pipeline around a structured table design. MCP lets the AI-assisted development environment work with Akeneo and Fabric through a single interface, explore the catalog structure, create the required tables, and troubleshoot the extraction.

The new pipeline captures changes through delta updates rather than rebuilding the full multilingual dataset. It runs automatically every morning, and a full export across all languages completes in approximately 2.5 to 3 hours without hitting concurrency limits.

MCP is not replacing Akeneo as the governed source of product information. It is helping DwyerOmega extend that trusted data into the company’s broader analytics and operational environment.

What comes next

DwyerOmega is testing richer model charts for products with extreme complexity, including one model with an estimated 100,000 to 200,000 possible permutations, and looking at ways to make Akeneo the governed source for AI agents that need accurate product context.

For Cyrus, the underlying value is structural.

The long-term opportunity is bigger than any single dashboard or automation. Instead of building another isolated integration for every new need, the company can create repeatable workflows around one trusted product foundation.

For other manufacturers, the lesson is practical: the value of governed product data grows when people and systems can use it without introducing another layer of inconsistency. DwyerOmega is showing how MCP can turn a PIM from a destination for product data into an active source of operational intelligence.

Where that foundation gets especially interesting is what DwyerOmega can now build on top of it. From giving teams greater visibility into product data quality to accelerating content creation and onboarding newly acquired product lines, MCP is beginning to change how product information moves across the business.

In Part 2, we’ll explore those workflows and the impact they’re already having.

The easiest way to build an AI platform is to keep telling yourself you are only building one agent at a time.

First, the content team creates an agent to generate product descriptions. Supplier operations builds another to map incoming data. Then eCommerce adds one to validate marketplace listings, while a different team experiments with enrichment or localization.

Each project solves a genuine problem. Each may be relatively straightforward to build. And none of them, on its own, looks like a platform.

But eventually, those agents need to work on the same products. They need to agree on which data is authoritative, apply the same business rules, complete tasks in the right order, and leave a record of what they changed and why. The architecture begins to take shape before anyone has consciously decided to build it.

That is why the build-versus-platform decision is not really about whether your engineering team can create an AI agent. In many cases, it can. The more important question is whether you are solving one contained problem, or creating the first part of a connected AI operating model that your team will need to govern, maintain, and expand over time.

When Building Your Own Agent Makes Sense

There are good reasons to build an AI agent internally.

A custom approach can work well when the use case is narrow, the inputs and expected outputs are clearly defined, and the underlying product information comes from a small number of reliable sources. It is also easier to manage when the catalog changes infrequently, only a few downstream systems are involved, and mistakes can be identified and corrected without significant business or regulatory impact.

In those circumstances, an internal agent can be a practical way to test an idea, solve a specific bottleneck, and build experience with AI.

The important question is ownership. Someone still needs to maintain the models, prompts, integrations, business rules, and monitoring as the technology and your requirements evolve. A custom build is most sustainable when the organization has both the engineering capacity and the long-term commitment to support it.

Where Complexity Begins to Compound

The challenge is rarely the agent itself. It is everything the agent depends on.

Imagine that a Supplier Data Agent updates a technical specification. Should the Content Agent immediately regenerate the product description? Does the Quality Agent need to validate the new value first? Should the Localization Agent translate the updated copy automatically? Does the Activation Agent pause publication until the change has been approved?

These are no longer model or prompting questions. They are questions about product data, business rules, workflows, accountability, and system design.

Without a shared operational layer, different agents may work from slightly different versions of the same product record. One may use an attribute value that another has already replaced. A Content Agent may generate copy before a validation step is complete. A Localization Agent may translate information that is still being revised upstream.

The result is a collection of useful automations that create delays, inconsistencies, and new points of failure when they need to work together.

What a Fleet of Agents Needs

As you move from one agent to several, four shared capabilities become increasingly important:

A shared product state

Every agent should read from and write to the same governed product record. When agents use separate copies, schemas, or data pipelines, inconsistencies can remain hidden until they cause a rejected listing, inaccurate description, or customer-facing error.

Sequencing and dependency management

Agents across the product lifecycle do not operate independently. Supplier information needs to be normalized before enrichment. Enriched data may need validation before content is created. Content may require approval before localization or activation. Those dependencies must be managed deliberately.

Shared business rules

Each agent makes decisions about what to complete, flag, publish, or translate. If the same rule is recreated across several prompts and scripts, a single business change may require updates in several different systems.

Unified observability

When something goes wrong, teams need to trace the complete chain of events. That means seeing the source information, the recommendation an agent made, the rules it applied, the person who reviewed it, and the final change made to the product record. Without that history, troubleshooting quickly becomes an investigation across systems and teams.

When a Platform Becomes the More Practical Choice

A platform becomes increasingly valuable when multiple agents need to share product data, business context, channel knowledge, approval workflows, and audit history.

It can also reduce the amount of work required to introduce each new use case. Instead of rebuilding governance, integrations, mappings, and monitoring every time, teams can reuse capabilities that are already in place.

That does not mean a platform removes the need for custom development. Many organizations will still build specialized agents, business logic, and experiences that reflect their specific products and processes.

The platform’s role is to provide the foundation around those agents: a governed product model, reusable rules, human workflows, channel connectivity, traceability, and the infrastructure required to keep everything aligned as the catalog evolves.

Ask What You Are Really Building

The build-versus-platform decision should not begin with, “Can our team build this agent?”

In many cases, the answer will be yes.

A more useful question is: “Are we building one contained capability, or the beginning of a connected system?”

For a stable, low-risk use case, an independent build may be entirely appropriate. But when several agents begin acting on the same products, supporting the same launches, and applying the same business rules, the surrounding operating layer becomes just as important as the agents themselves.

Start with a genuine operational problem, use a representative sample of your product data, define where human judgment is required, and measure the new process against the one it replaces. That will help you determine whether you need a focused custom solution today, or a shared foundation that can support the next use case tomorrow.

Amazon listing problems have a habit of appearing at the worst possible moment.

A new product is ready to launch, but required attributes are missing. A variation suddenly appears as a separate listing. A specification is correct on your website but outdated on Amazon. Or a team discovers that hundreds of SKUs need to be manually updated because a marketplace requirement has changed.

It is tempting to treat each of these as an Amazon problem.

But in many cases, Amazon is simply where an upstream product information problem becomes visible.

If the product data feeding Amazon is incomplete, inconsistent, incorrectly structured, or scattered across different systems, spreadsheets, and teams, no amount of last-minute listing optimization can fully solve the problem. The more products and marketplaces you manage, the more important it becomes to fix product information at the source.

What Causes Amazon Listing Data Problems?

Amazon listing data problems often originate from poor-quality product information before that information ever reaches the marketplace. Common causes include:

• Missing attributes 
• Inconsistent values 
• Incorrect product relationships 
• Outdated specifications 
• Fragmented supplier data 
• Manual channel-specific transformations

Here are some of the problems brands and retailers should watch for.

1. Missing or incomplete product attributes

A product may technically have enough information to exist in an ERP or internal database while still lacking the information necessary to create a useful marketplace experience.

Dimensions, materials, compatibility details, ingredients, technical specifications, care instructions, certifications, or other attributes may be blank or buried inside unstructured descriptions.

That creates problems at two levels. First, marketplace requirements may make some fields necessary for successful publication. Second, even when a listing can go live, missing information gives customers less confidence that a product meets their needs.

That matters because shoppers are actively scrutinizing products; 29% of consumers review detailed product information when assessing deals specifically on Amazon, while only 9% trust deals without verifying them first.

2. Product data that does not match Amazon’s structure

Every sales channel has its own way of organizing product information.

Your internal attribute might be called “fabric,” while a marketplace expects “material.” A technical measurement may need a particular unit. A value stored as free text internally may need to map to a predefined marketplace field.

When these differences are handled manually every time products are published, inconsistencies multiply quickly.

The better approach is to maintain structured, governed product information centrally and establish repeatable mappings that transform that data into Amazon’s required format. This separates the underlying product truth from the way each individual channel needs to receive it.

3. Broken parent-child and variation relationships

Color. Size. Pack quantity. Finish. Configuration.

Variants are useful to customers, but only when their underlying relationships are modeled correctly.

If parent-child relationships are inconsistent upstream, variants can appear as disconnected products rather than intuitive choices within the same product experience. Shoppers may struggle to understand their options, while commerce teams spend time manually correcting relationships downstream.

This is fundamentally a product modeling problem. Managing product families and variations centrally makes it much easier to maintain consistent relationships as catalogs expand.

4. Conflicting or outdated product information

A product may be listed as 50 cm wide on your own website, 52 cm on Amazon, and 51 cm in a distributor catalog.

Which answer should the shopper trust?

These discrepancies frequently arise when information is stored across ERP platforms, spreadsheets, supplier files, eCommerce systems, and marketplace tools without a governed source of product information.

And shoppers increasingly have ways to spot those inconsistencies. 62% of U.S. consumers compared prices across retailers during Amazon’s 2026 Prime Day. When consumers move easily between channels, conflicting information becomes much harder to hide.

5. Weak content created from weak source data

Compelling Amazon titles, descriptions, bullet points, and rich content ultimately depend on the quality of the product information behind them.

If a marketing team starts with little more than a product code, a generic supplier description, and a handful of specifications, it becomes much harder to create content that clearly communicates what makes the product useful, distinctive, or relevant to a particular buyer.

The same challenge applies when generative AI is used for enrichment and content creation. AI can help teams generate descriptions, translate content, classify products, and adapt messaging for different channels, but it still relies on accurate and complete product context. Missing attributes, vague specifications, or inconsistent source data can quickly turn into equally vague or inaccurate generated content.

That makes strong source data increasingly important as brands scale marketplace content. Structured attributes, approved terminology, trusted specifications, and clear product relationships give both people and AI better material to work with, making it easier to create Amazon content that is accurate, differentiated, and consistent across large catalogs.

6. Marketplace growth that depends on manual work

A few Amazon listings can be managed manually.

A few thousand listings, constantly changing across markets, product categories, suppliers, and channels, are a different story.

Manual spreadsheets, copying and pasting, and one-off data transformations create opportunities for errors while making every catalog update slower.

Synchro Diffusion offers a good example of what happens when marketplace teams tackle both enrichment and activation together. The company used Akeneo’s native AI capabilities to automatically generate product descriptions, translate content, and structure product data, increasing data completeness from 40% to 90%. It then used Akeneo Activation to connect that enriched product information directly to Amazon, giving the team greater control over product data, pricing, and orders while making it easier to resolve issues at the source and publish updates faster.

Fix Amazon Listing Problems at the Source

Teams struggling with recurring Amazon listing issues should ask a different question.

Instead of “How do we fix this listing?”, ask “Why did incorrect or incomplete information reach this listing in the first place?”

That means establishing a centralized product information foundation, defining required attributes by product category, improving supplier data collection, governing variant relationships, validating completeness before publication, and creating reusable mappings for channel requirements.

A Product Information Management (PIM) system can provide that foundation by giving teams one governed environment where product data can be collected, enriched, validated, and prepared for marketplaces. Syndication and activation capabilities can then transform that trusted information into the formats individual channels require.

Because the most expensive Amazon listing problem is rarely one bad field. It is discovering that the same bad field exists across hundreds or thousands of products.

Fix the product information foundation, and Amazon becomes another channel where trusted, complete, compelling product experiences can be activated at scale.

For years, product information management could be treated primarily as an operational discipline: centralize the catalog, improve data quality, and make sure accurate information reaches the right channels.

Recently, that job has become much bigger.

Product information now supports everything from eCommerce and marketplaces to internal operations, partner ecosystems, customer experience, and increasingly, AI-powered discovery and automation. As those environments become more connected, enterprises need product information that can move between systems and teams without losing the governance, context, and control the business depends on.

In other words, product data is becoming infrastructure for how modern businesses operate, engage customers, and prepare for AI.

That changing role is reflected in two recent evaluations published by ISG Software Research: the ISG Buyers Guide™ for Product Information Management Platforms 2026 and the ISG Buyers Guide™ for Product Experience Management 2026.

ISG Research defines PIM around managing and using product information across front- and back-office applications and processes. PXM extends that focus to managing customer experiences through systems that support product-related activities.

Together, the two guides reinforce an increasingly important point: modern PIM and PXM decisions are also decisions about integration, governance, automation, and AI readiness.

Modern PIM has to connect the enterprise

The ISG Buyers Guide™ for Product Information Management Platforms 2026 evaluated 21 software providers. Akeneo was placed in the Exemplary category, while the Buyers Guide Insight highlighted strengths across both core platform requirements and emerging AI capabilities.

ISG Research specifically highlighted Akeneo’s performance in Integration, an increasingly important consideration when product information has to flow between PIM, commerce platforms, ERP systems, supplier ecosystems, marketplaces, and customer-facing channels. It also identified License, Use and Audit as an area of differentiation, alongside Privacy and Technology Administration capabilities that support enterprise oversight.

When product data touches more systems, processes, teams, and automated workflows, enterprises need to know that information can move without sacrificing control or transparency.

At the same time, ISG Research highlighted Akeneo’s capabilities in Generative AI and Platform Support for AI, connecting the traditional role of PIM with a rapidly emerging enterprise requirement: preparing product information for AI-enabled use cases.

View the Akeno ISG Buyers Guide Insight for PIM Platforms here.

PXM turns the foundation into an experience

Managing product information, however, is only part of the equation.

Customers never see your taxonomy or data governance model. They see the outcome: the description on a product page, the specifications on a marketplace, the search result that helps them discover an item, or the answer an AI assistant gives when they ask for a recommendation.

That is where PXM comes in.

The ISG Buyers Guide™ for Product Experience Management 2026 evaluated 14 software providers, and Akeneo again was identified as an Exemplary Provider.

ISG Research highlighted Akeneo’s performance across several areas, with what it described as its most pronounced advantages in Generative AI and Intelligent Automation. Integration, Privacy, Technology Administration, and Investment were also called out in the Buyers Guide Insight. That combination matters as product experience becomes more complex.

Teams are being asked to manage expanding catalogs, more channels, more content variations, localization, changing marketplace requirements, and increasingly personalized experiences. Intelligent automation can help teams handle that complexity more efficiently, but only when the information underneath those workflows is reliable and connected.

View Akeneo’s ISG Buyers Guide Insight for Product Experience Management here.

AI readiness starts with product data readiness

This may be the most important connection between the two Buyers Guides.

AI-powered commerce introduces a new consumer of product information: the machine.

Generative AI applications, conversational interfaces, intelligent automation, AI-powered search, and emerging autonomous agents all depend on the information available to them. If that product information is fragmented, inconsistent, inaccessible, or poorly governed, adding AI does not remove the underlying problem.

It can simply make the problem move faster.

That means AI readiness cannot be separated from product data readiness.

Before organizations can scale AI-powered product experiences, they need confidence in the foundation underneath them: accurate information, integration across the enterprise, appropriate governance, operational oversight, and workflows capable of supporting automation.

The two ISG evaluations reflect that changing requirement. Across its assessments of Akeneo, ISG Research highlights capabilities spanning Integration, Generative AI, Platform Support for AI, Intelligent Automation, Privacy, and Technology Administration.

Product information is becoming enterprise infrastructure

Enterprises increasingly need to govern product information across the organization and transform it into accurate, compelling experiences across every customer touchpoint. Treating those as separate challenges makes it harder to create the connected foundation that modern commerce (and increasingly AI) depends on.

That is the role Akeneo Product Cloud is designed to play: helping organizations make product information governed, connected, intelligent, and ready to support experiences across channels, systems, and business processes.

Because as AI becomes more deeply embedded in how products are managed, discovered, evaluated, and purchased, one principle becomes increasingly difficult to ignore:

The next generation of product experiences will only be as reliable as the product information behind them.

If you work with product data, there is a good chance you’ve already implemented your first AI use case.

Maybe you want to automate supplier onboarding. Perhaps your team is experimenting with AI-generated product descriptions. Maybe enrichment is taking too long, localization is creating bottlenecks, or channel errors are eating up hours every week, and so you’ve been testing out how AI can help optimize these processes, and maybe you’ve already seen success with some of these programs.

But once one AI-powered workflow starts delivering value, the natural question becomes: what else could we automate?

That is where product teams begin moving from a single AI assistant toward something much more interesting: a collection of specialized AI agents working across the product lifecycle.

There is no single “AI agent for PIM.” There are many potential agents, each solving a different problem and supporting a different stage of product operations. The ten below represent some of the highest-impact opportunities product data teams should be thinking about today.

1. Product Modeling Agent

Before a product can be enriched, marketed, or published, it needs the right structure.

A Product Modeling Agent can analyze product families, variants, relationships, and taxonomies to identify inconsistencies, duplicates, or opportunities to improve how a catalog is organized.

Think of it as an always-on partner for keeping your product model aligned with the reality of what you sell.

2. Supplier Data Agent

Supplier onboarding is still one of the most manual parts of product information management.

A Supplier Data Agent can connect to supplier sources, extract and map product information, normalize formats, validate incoming data, and prepare products for enrichment. Instead of teams spending hours reconciling spreadsheets and files, the agent handles much of the repetitive groundwork.

3. Enrichment Agent

This is one of the most obvious AI opportunities for product teams.

An Enrichment Agent identifies missing or incomplete product information and fills gaps using trusted sources such as supplier documentation, technical PDFs, imagery, or approved web content.

The important word there is trusted. AI does not magically fix fragmented data. If the underlying sources conflict, an agent can just as easily scale the wrong answer as the right one.

4. Content Agent

A Content Agent turns structured product information into customer-facing content.

Descriptions, titles, bullet points, FAQs, marketing copy, and channel-ready content can all be generated from the same governed product record while following approved terminology and brand guidelines.

The biggest opportunity here is producing content that is consistent across thousands of products without requiring your team to start every page from scratch.

5. Localization Agent

Translation is only one part of localization.

A Localization Agent can adapt content for language, terminology, cultural expectations, and market-specific requirements. The goal is not to produce a literal translation, but content that is ready for a specific market.

That becomes increasingly valuable when product teams are managing many regions, languages, and local requirements at once.

6. Asset Intelligence Agent

Product assets are often treated separately from product information, even though the two are deeply connected.

An Asset Intelligence Agent can analyze product images and other digital assets, generate metadata, assess visual quality, connect assets to the correct products, and flag where additional imagery or information may be needed.

That can significantly reduce the manual work involved in organizing and validating large asset libraries.

7. Quality & Compliance Agent

The bigger your catalog becomes, the harder it is for people to spot every inconsistency.

A Quality & Compliance Agent can continuously monitor catalog health, identify incomplete or potentially incorrect information, check business and regulatory rules, and recommend corrective actions.

Crucially, it won’t make every decision itself. High-risk or uncertain recommendations can (and should) still be routed to a human expert for review.

8. Activation Agent

Creating good product information is only useful if channels can actually accept it.

An Activation Agent monitors connected marketplaces, retailers, and other destinations for publication errors, changing requirements, and policy issues. It can identify root causes, group similar problems, and recommend or prepare corrective actions.

Instead of discovering errors after a listing fails, teams can move toward proactively managing channel readiness.

9. Market Intelligence Agent

Most product data workflows still operate primarily in one direction: create the information, publish it, and move on.

A Market Intelligence Agent closes that loop.

It can monitor channel performance, reviews, customer behavior, and AI search visibility, then feed those signals back into the product information process so teams know what should be improved next.

10. Insights Agent

The final agent looks across the system rather than focusing on one specific task.

An Insights Agent can analyze catalog performance, customer behavior, channel signals, and AI discovery data to identify opportunities, recommend actions, and measure business impact over time.

In other words, it helps product teams move from managing tasks to making better decisions.

A Note on Human Oversight

None of these agents should be thought of as fully autonomous replacements for product data teams.

Their real value is in taking on repetitive, manual, and time-consuming work such as classifying supplier data, identifying missing attributes, generating first drafts, checking for inconsistencies, or preparing content for activation, all while keeping people in control of the decisions that require judgment.

That means human review still very much matters. Product information can carry legal, regulatory, technical, commercial, and brand implications, so teams need clear boundaries around what an agent can complete automatically, what should be recommended for approval, and what must remain human-led.

The goal is not to remove people from the process. It is to make better use of their time.

When agents handle the predictable work and surface the exceptions that deserve attention, experts can spend less time on repetitive tasks and more time applying the context, experience, and judgment that AI cannot provide on its own.

The Future of PIM Agents

You do not need to build all ten of these tomorrow, nor can you.

Most product teams will start with one or two high-impact use cases: perhaps supplier onboarding, enrichment, content generation, or activation.

But there is a bigger idea worth planning for.

As individual AI capabilities become easier to access, the long-term differentiator will not be having a single Content Agent or an Enrichment Agent, but will come from how effectively multiple agents work together and interact with each other.

Does your Content Agent know when the Enrichment Agent has updated an attribute? Does your Localization Agent wait until content has been approved? Does your Activation Agent understand which version of the product record is authoritative?

That is where trusted product data, shared business rules, human oversight, channel intelligence, and traceability become essential. Without that foundation, a collection of useful AI tools can quickly become a collection of automation silos.

The best place to start is with the use case that solves a real operational problem today, and start building a foundation that makes the next agent easier, safer, and more valuable to introduce tomorrow.

For the past few years, much of the conversation around enterprise AI has focused on the model.

Which model should we use? How powerful is it? How accurate are its answers? Can it generate better content, automate a workflow, or power an AI agent?

But as organizations move from experimenting with AI to actually putting it to work, another question is becoming much more important:

What are the guardrails of our AI model?

An AI model on its own may be able to generate a convincing answer, but it does not automatically know which product information it is allowed to use, which business rules it needs to follow, when a human needs to approve an action, or whether it has permission to update a system.

That surrounding infrastructure is increasingly being described as the AI harness.

What is An AI Harness?

An AI harness is the combination of systems, controls, context, and workflows that allow an AI model or agent to operate reliably within a business.

Think of the model as the engine. The harness is everything that connects that engine to the rest of the vehicle and keeps it moving in the right direction.

It can include capabilities such as:

• Memory and context that give AI the information it needs to make relevant decisions
• Permissions that determine what information an agent can access and what actions it can take
• Guardrails and business rules that define acceptable behavior
• Orchestration that coordinates AI across systems, workflows, and other agents
• Human approval workflows for decisions that require expertise or oversight
• Auditability and traceability so organizations can understand what happened, why, and using which information

This idea becomes especially important as businesses move toward agentic AI.

While a chatbot can really only answer a question, an autonomous AI agent may actually take an action by enriching a product record, translating product copy, classifying an item, publishing information to a marketplace, or recommending a pricing change.

Once AI moves from answering to acting, the infrastructure around it becomes critical.

Why AI Needs More Than Good Prompts

Early AI experimentation often starts with prompts.

Give a model the right instructions, provide some information, refine the prompt, and see what happens.

That works well for experimentation. It becomes much less dependable when AI is connected to operational processes.

If you have an agent tasked with improving product descriptions, then it needs to understand questions like:

• Which product attributes are approved sources of truth? 
• Which claims can legally appear in a particular market? 
• What tone should the brand use? 
• Which terminology is prohibited? 
• Does the description need human review? 
• Which version should be published to Amazon versus the company’s ecommerce site?

These questions require access to trusted systems, structured product information, business context, permissions, governance, and workflows.

In other words, they require a harness.

The AI Harness Connects AI to Systems of Record

For many enterprises, the foundation for an AI harness likely already exists.

ERP systems contain operational information, PIM platforms govern product information, DAM systems manage digital assets, and commerce platforms, marketplaces, supplier systems, and customer-facing applications each hold additional context.

The challenge is enabling AI to interact with these systems in a controlled way.

For product-related AI in particular, a PIM can play an important role as a trusted system of record. Instead of allowing an agent to pull product information indiscriminately from spreadsheets, websites, supplier files, and disconnected databases, organizations can give it governed access to consistent, validated product data.

That dramatically changes what AI can do.

What Does an AI Harness Look Like inPractice?

There is no single piece of software called an AI harness that every organization can simply install. Instead, businesses will build the capabilities progressively around the AI use cases they deploy.

A practical implementation can begin with four areas:

1. Establish trusted data foundations

AI can only work reliably with the information available to it.

Before giving agents greater autonomy, organizations need to identify the systems that contain authoritative product information and reduce fragmentation between supplier data, ERP records, PIM systems, commerce platforms, and channel feeds.

Structured, complete, governed product information gives the harness a dependable foundation.

2. Give AI business context

An AI agent needs to understand how your organization operates: taxonomies, approval requirements, channel specifications, brand terminology, regulatory rules, localization standards, and other repeatable decisions.

Capturing that knowledge as reusable context allows AI to operate according to the business rather than generating generic outputs.

3. Define permissions and human checkpoints

Not every AI action should have the same level of autonomy. An agent might be allowed to automatically fill a low-risk missing attribute but require approval before changing a compliance claim or publishing information to a retailer.

Organizations should explicitly define what AI can read, recommend, change, approve, and publish.

Human-in-the-loop workflows then become part of the harness rather than an emergency safeguard added after something goes wrong.

4. Build traceability into every action

As AI takes on more operational responsibility, businesses need to know where its decisions came from.

Which source data did the agent use? What instruction was it given? Which rule was applied? Did a human approve the result? What was eventually published?

That audit trail becomes essential for governance, troubleshooting, compliance, and ultimately trust.

The Real AI Advantage is What You Build Around It

The race to adopt AI can make it feel like the model is the main event. But models are becoming more powerful, more accessible, and increasingly interchangeable.

The harder thing to replicate is the environment you create around them.

Your product knowledge. Your business rules. Your approval processes. Your channel requirements. Your governance. Your understanding of what AI should be allowed to do independently and where human judgment still matters.

So instead of asking whether your organization has access to the latest AI, ask a more revealing question:

If you gave an AI agent the keys to part of your product experience today, would it know how to drive?

The AI harness is what makes the answer yes

My sister recently needed to replace her child’s car seat before a long road trip. 

She needed one that fit her particular car model, worked for a specific age and weight range, met the right safety standards, had a washable cover, and could be delivered within two days, in time for the road trip.

A few years ago, that would have meant opening a handful of retailer tabs, checking specifications, reading reviews, comparing delivery dates, and probably second-guessing whether it was actually compatible with her car model or would actually arrive on time. 

Instead, she went straight to ChatGPT.

She described exactly what she needed and asked it to narrow down the options. Within minutes, she had a shortlist of products that appeared to meet her criteria, along with explanations of why each one might be a good fit.

That experience captures a much bigger shift in how people are beginning to shop. Increasingly, consumers are not starting with a search bar and working their way through a list of links. They are starting with a question, a problem, or a set of constraints and asking AI to do more of the research for them.

And this is only the beginning.

As AI evolves from recommending products to acting on behalf of shoppers, we are moving toward agentic commerce: a world where AI agents can research options, compare products, assess details like price, availability and policies, and potentially complete purchases with far less human intervention.

That changes the rules of product discovery. AI agents need product information they can actually interpret, verify, compare, and trust.

That is where Answer Engine Optimization, or AEO, comes in. And more importantly, it is why preparing your product data for AI today is quickly becoming one of the most important steps toward succeeding in agentic commerce tomorrow.

What Exactly is Agentic Commerce?

Today, most AI-powered shopping experiences still involve a human somewhere in the decision-making process. A shopper asks a question, an AI assistant recommends products, and the shopper decides what to do next.

Agentic commerce pushes more of that journey onto the AI.

An agent might interpret a request, establish requirements, compare products, evaluate compatibility, check delivery or return policies, assess trust signals, and eventually complete a transaction. 

The next phase of AI-powered discovery happens when agents move beyond answering questions to evaluating and potentially acting on behalf of shoppers.

But an agent can only make decisions based on the information available to it.

If your dimensions are missing, compatibility information is buried in a PDF, availability is outdated, or two retailers list conflicting specifications, the agent has a problem. It cannot confidently determine whether your product meets the shopper’s requirements.

And confidence matters when the AI is deciding whether to recommend or purchase it.

Where Does AEO Fit In?

SEO was designed for a world of keywords, rankings, traffic, and clicks. The objective was to get a page high enough in search results that a human would click through, visit your website, and continue their journey there.

AEO focuses on a different outcome: making information understandable, credible, relevant, and structured enough that an AI engine can use it in an answer or recommendation.

That distinction is increasingly important because SEO success does not automatically equal AI visibility. A product can rank well in conventional search while remaining difficult for AI to recommend because its underlying information is incomplete, inconsistent, unstructured, or difficult to verify.

Ultimately, AEO (and agentic commerce readiness) comes back to the product data itself.

How to Prepare Product Data for Agentic Commerce

1. Start with a governed source of truth

AI readiness becomes much harder when product information is scattered across spreadsheets, ERP systems, supplier portals, eCommerce platforms, and regional teams.

Agents need reliable product information, and organizations need a clear place where that information can be governed.

That is already influencing AI architecture; 65% of organizations rely on structured systems such as PIM or ERP to power their AI models, partly in response to integration challenges and data silos.

A centralized product information foundation gives teams somewhere to standardize, validate, enrich, approve, and distribute the information agents will depend on.

2. Make product information machine-readable

Beautiful product pages are designed for people. Agents need information designed for machines.

That means structured attributes, standardized categories, global identifiers, relationships between products, and appropriate product schema. These elements create the translation layer between human-facing content and machine interpretation.

Instead of forcing an AI to extract a product’s dimensions from a paragraph, give it clearly defined height, width, depth, material, compatibility, and other relevant attributes.

The less an agent has to infer, the easier your product is to evaluate.

3. Think beyond traditional completeness

A completed product record should not simply mean that every mandatory field in your PIM is filled in. Completeness increasingly needs to reflect buyer intent.

If someone asks for “a waterproof backpack suitable for commuting with a 16-inch laptop,” does your product data explicitly contain enough information to establish waterproofing, laptop compatibility, capacity, intended use, and relevant materials?

That’s why we recommend enriching records around environments, occasions, personas, problems, compatibility, and customer intent so AI can match products to real-world situations.

4. Replace vague language with useful facts

AI systems need evidence.

“Sleek design” tells an agent very little. “6-inch footprint that fits beneath standard kitchen cabinets” is far more useful. Specific, extractable facts help AI systems connect product characteristics with what a shopper is actually asking for.

Look at reviews, customer questions, search queries, and return reasons too. They can reveal the information shoppers care about that may never have made it into the original product record.

5. Keep information consistent everywhere

An agent may encounter your product on your website, a retailer site, a marketplace, a product feed, a review platform, or somewhere else entirely.

Conflicting prices, specifications, identifiers, or claims create uncertainty. The goal should be one governed product record that distributes consistent, enriched information across every relevant channel.

The Real Work of Agentic Commerce Happens Before the Agent Arrives

With the growing popularity of AI-powered search and discovery, a shopper may never read your carefully written description, scroll through your comparison table, or click through your category navigation. An AI agent may encounter your product somewhere else entirely, pull together what it can find, compare it against a dozen alternatives, and make a judgment in seconds.

In that moment, your product data becomes your salesperson.

It has to explain what the product is, who it is for, what it works with, why it is different, whether it is available, and whether the information can be trusted. There is no opportunity for a merchandising team to step in and clarify a missing attribute. No second chance to explain that an outdated retailer listing is wrong. The agent works with the evidence it has.

That is why preparing for agentic commerce is less about chasing the newest AI capability and more about making your catalog impossible to misunderstand.

For the past few years, the Digital Product Passport (DPP) has been easy to file under “something we’ll need to worry about eventually.”

Eventually is now getting a lot closer.

The infrastructure is now being put in place, product-specific rules are moving forward, and by February 2027, certain batteries sold in the EU will need a digital passport containing detailed, accessible information about everything from their composition and performance to their origins and end-of-life handling.

And that first deadline matters even if you do not sell batteries.

The Digital Product Passport is part of a much broader shift in how the EU expects companies to manage and communicate product information. Over time, more manufacturers, retailers, distributors, and brands will need to prove not only what a product is, but where its materials came from, what claims can be substantiated, how it should be repaired or recycled, and whether that information can be traced back to a reliable source.

That makes DPP compliance much more than a QR-code project. It is fundamentally a product information challenge.

So, what is actually mandatory today? What changes in 2027? And what should companies be doing now to make sure their product data is ready?

What is a Digital Product Passport?

A Digital Product Passport is a structured digital record associated with a product, component, or material. It is designed to make reliable information available throughout the product lifecycle, supporting sustainability, circularity, and regulatory compliance.

Depending on the category, a DPP may contain information about product identity, materials, origin, environmental performance, durability, repair, reuse, hazardous substances, and recycling. Users access it through a data carrier, such as a QR code, attached to the product, packaging, or accompanying documentation.

A DPP must use a unique identifier, follow applicable interoperability requirements, be registered in the EU DPP Registry, and remain accurate over time. The Registry stores identifiers and metadata, while detailed data remains decentralized under the responsible operator or a service provider.

Which companies are affected?

The Ecodesign for Sustainable Products Regulation (ESPR) establishes a framework that can ultimately apply to most physical goods placed on the EU market, including components and intermediate products. Food, feed, medicines, living organisms, and products of human origin are among its exclusions.

The ESPR does not automatically require every covered product to have a passport. Product-specific delegated acts determine which products need one, what information it must include, and when compliance begins.

The rules apply to domestic and imported products. Responsibility can fall on manufacturers, authorized representatives, importers, distributors, dealers, or fulfillment service providers. Primary responsibility for creating an accurate DPP generally sits with the economic operator placing the product on the EU market. Online marketplaces will also need to make applicable passports accessible during distance sales.

A UK, US, or Asian brand is therefore not outside the regime simply because it has no EU factory. 

Any business selling an in-scope product into the EU must determine which operator is responsible and ensure the required information can move across the supply chain.

What is mandatory in 2026?

For most sectors, 2026 is the year the DPP infrastructure becomes operational, not the year every product needs a passport.

The EU DPP Registry became operational on July 20, 2026. Six harmonized technical standards were formally adopted in July, with two additional standards scheduled for September. The Commission also plans to adopt product-specific requirements for iron and steel in the fourth quarter of 2026.

Yet an adoption date is not a compliance date. The Commission states that companies will receive a transition period of at least 18 months after an ESPR delegated act is adopted. On that minimum timetable, even if the iron-and-steel act arrives as planned in late 2026, its practical deadline would fall in 2028 or later.

Inclusion in the ESPR working plan means a product group is being assessed and regulated; it does not create an immediate DPP obligation.

What becomes mandatory in 2027?

The clearest binding deadline is February 18, 2027. From that date, every electric-vehicle battery, light-means-of-transport battery, and industrial battery with a capacity greater than 2 kWh placed on the EU market or put into service must have a battery passport.

This covers EV batteries, batteries for e-bikes, e-mopeds, and e-scooters, and qualifying industrial batteries. The operator placing the finished battery on the market (not the supplier of an individual cell or module) is responsible for creating and maintaining the passport. It must be linked to the battery through a QR code.

The Commission’s August 2026 guidance maps 71 possible battery-passport data points and identifies which are mandatory, optional, conditional, or not yet required at launch. Relevant information includes identifiers, manufacturer details, model and serial or batch information, manufacturing location and date, weight, capacity, chemistry, critical raw materials, hazardous substances, performance, durability, and information supporting repair, reuse, and recycling.

Other sectors will advance during 2027. Delegated acts are planned for construction materials, textiles, aluminum, and tires. These are expected rule-adoption milestones, not automatic 2027 compliance deadlines. A planned 2027 textiles act, for example, does not mean every garment must carry a DPP before year-end; the minimum transition period still applies.

What is required for DPP success?

The visible QR code for a DPP is actually the final step. The difficult work is creating complete, transparent, and traceable product information behind it.

Completeness means identifying every required field, determining whether it applies at model, batch, or item level, and collecting it from systems and suppliers. Data may be spread across PIM, ERP, PLM, compliance, lifecycle-assessment, and supplier platforms. Missing evidence cannot be solved with polished copy at the last moment.

Transparency means making reliable information available to the right audience. Some data will be public, while other information may be restricted to authorities, repairers, or authorized actors. Sustainability claims should connect to calculations, certificates, and source records, not stand as unsupported marketing statements.

Traceability means giving products and operators stable identifiers and preserving the origin, timestamp, ownership, and approval history of each data point. Passport information must remain accurate as a product is repaired, repurposed, or otherwise changes.

Companies also need named data owners, validation workflows, supplier requirements, change controls, and interoperable systems. A central product information layer can coordinate the data, but it must connect to the specialist systems where technical, environmental, and compliance evidence originates.

The 2027 battery deadline is only the beginning. As DPP requirements expand into additional sectors, businesses will need to have a reliable foundation for managing complete, transparent, and traceable product information.

That means bringing product data together, improving its quality, establishing clear governance, and making it easier to connect information from suppliers, compliance systems, ERP, PLM, and other sources. Those capabilities are valuable for DPP compliance, but they also support better product experiences, faster channel activation, and stronger readiness for future regulatory requirements.

Akeneo can help businesses build that foundation. By centralizing, enriching, governing, and distributing trusted product information, Akeneo gives teams a more scalable way to prepare for evolving requirements like the Digital Product Passport.

See how Akeneo can help you build a product information foundation that is ready for what comes next.

For years, digital visibility has largely meant one thing: search engine optimization (SEO). We optimized product pages around keywords, technical performance, and search rankings with the goal of getting shoppers to click through to our websites.

That strategy still matters. But the way people discover products is changing.

Instead of searching for “best coffee machines” and scrolling through pages of results, a shopper can now ask ChatGPT, “What’s the best coffee machine for a small apartment that can make espresso?” ChatGPT then interprets the shopper’s intent, compares different products, evaluates available information, and produces an answer.

That shift is where Answer Engine Optimization (AEO) comes in. AEO is the practice of structuring product content and building brand authority so that AI engines can confidently cite and recommend your products.

The important distinction is that SEO primarily focuses on optimizing pages to rank in traditional search, while AEO focuses on whether your products are visible and understandable within AI-generated answers. A product can rank well on Google and still remain virtually invisible to an AI system if its product information is incomplete, inconsistent, difficult to verify, or inaccessible in a machine-readable format; we call this the Invisible Shelf.

So how do you make your products easier for AI to understand and recommend? It starts with four pillars: Structure, Authority, Freshness, and Relevance.

1. Structure: Be Understandable

Before an AI engine can recommend your product, it has to understand what it actually is.

AI engines rely on consistent categorization and clearly defined attributes to connect products with shopper intent. Structured data, including Schema.org markup, essentially acts as a translator between the product information humans see on a page and the machine-readable facts AI systems can interpret.

Let’s go back to our shopper looking for a coffee machine for a small kitchen. An AI system may interpret that request as a combination of specific requirements: an espresso machine, a compact footprint, and suitability for a countertop. If dimensions, product type, or other relevant attributes exist only in a paragraph of marketing copy (or are missing altogether) the AI may struggle to make that connection.

The lesson is simple: do not make AI guess. Product dimensions, materials, compatibility, identifiers, features, and other important facts should be clearly structured and accessible.

2. Authority: Be Credible

Being understandable is only the first hurdle. AI also needs to trust what it finds.

AI engines can cross-reference product and brand information across multiple independent sources before deciding whether a recommendation is credible. Your own website may serve as the primary source of truth, but information from knowledge bases, business databases, retailers, reviews, community platforms, and media coverage can all contribute to the broader picture.

Imagine your website says a product has one set of specifications while a major marketplace lists something different. A human shopper might simply assume one page has not been updated. For an AI system trying to determine which information is accurate, however, that contradiction creates uncertainty.

That is why consistency across channels matters so much. If product descriptions, specifications, identifiers, policies, or other details differ from one destination to another, confidence can erode. A governed source of truth that distributes consistent product information across channels gives AI systems a much clearer foundation to work from.

3. Freshness: Be Relevant

Real-time data velocity is an important trust signal for AI engines. Products need active, recent information around them if they are going to maintain visibility over time.

For your brand, that might mean recent customer sentiment on review platforms or up-to-date signals from major marketplaces. In B2B environments, freshness can come from mentions in analyst reports, partner ecosystems, professional directories, or industry publications.

Maybe you have a product that launched three years ago and has since been updated with new functionality. If most third-party descriptions still reflect the original version, an AI engine may encounter conflicting evidence about what the product can actually do.

AEO therefore cannot be treated as a one-time optimization project. Product information, marketplace content, reviews, and external references all need ongoing attention as products and markets evolve.

4. Relevance: Be Recommended

The final pillar may be the most important distinction between simply being visible and actually being chosen.

Technical completeness makes your product understandable. Context makes it recommendable. AI engines increasingly respond to conversational questions built around needs, situations, and use cases rather than simple keyword combinations.

That means product content should go beyond generic claims like “sleek design” or “high performance.” Specific, extractable facts are far more useful. 

Context also means connecting products to real-world occasions and needs. A coffee machine is not always just a coffee machine. Depending on its attributes, it might also be a compact espresso maker for a studio apartment, a housewarming gift under £100, or an everyday appliance for a remote worker. Those details help AI connect a product to the way someone is actually asking a question.

You can go one step further by providing clear answers to common customer questions directly within product content. AI engines are designed to synthesize answers, so making useful information easy to extract increases the likelihood that your product can become part of that answer.

From Search Visibility to AI Readiness

The four pillars of AEO work together: structure helps AI understand your products, while authority helps it trust them. Freshness shows that the information remains relevant, and context helps AI connect the product to a shopper’s specific need.

Ultimately, AEO is not about abandoning SEO or chasing another marketing acronym. It reflects a broader change in product discovery. As shoppers increasingly rely on AI to compare options and make recommendations, product information has to work for machines as well as humans.

For organizations running SAP Commerce on-premise (formerly SAP Hybris Commerce), July 31, 2026 was an important planning date.

On that day, version 2205, the final on-premise release, reached End of Mainstream Maintenance and moved into Customer-Specific Maintenance.

It’s important to note that SAP has stated that there is currently no defined sunset or end-of-life date for SAP Commerce on-premise, and this change does not affect SAP Commerce Cloud. 

The other good news is that this is not an overnight shutdown, nor does it automatically require every organization to replace its product information management capabilities. It is, however, a valuable moment to review the role SAP Commerce plays in your technology landscape and decide what should come next.

A modernization milestone, not a forced migration

When SAP Commerce 2205 enters Customer-Specific Maintenance, customers can continue to access support under their existing arrangements, but the scope changes. SAP will no longer adapt the release to certain new or changing requirements, including legal changes, technological updates, new support packages, updates to third-party libraries, or support for new interfaces.

For some organizations, the logical next step will be a move to SAP Commerce Cloud. Others may use the milestone to compare commerce platforms, adopt a more composable architecture, or modernize several parts of their digital stack at once.

There is another important point to clarify: Product Content Management, or PCM, is not a standalone SAP product that is independently reaching end of life. It is an embedded capability within SAP Commerce, and SAP Commerce Cloud continues to include product, content, and catalog management capabilities. An organization moving to SAP Commerce Cloud can therefore continue using SAP’s embedded PCM without adopting a separate PIM.

The right conversation is not, “What must we replace?” It is, “What architecture will best support the business we are becoming?”

Modernization is about more than infrastructure

Moving from customer-managed infrastructure to a cloud service can be a major undertaking, but infrastructure is only one part of modernization.

Product information now supports far more than a single eCommerce storefront. The same data may need to power regional sites, mobile experiences, marketplaces, distributor portals, print catalogs, customer service tools, retail partners, and emerging AI-powered discovery channels. It must also serve teams across marketing, eCommerce, product, compliance, sales, and operations.

At the same time, organizations are investing in AI and composable commerce. Both increase the importance of trusted, structured, well-governed product information. An AI agent cannot reliably compare, describe, or recommend products when attributes are incomplete, terminology is inconsistent, or critical context is scattered across systems. A composable architecture also delivers limited flexibility when product data remains tightly coupled to one commerce platform.

That makes this milestone an opportunity to ask a broader strategic question: Should product information remain embedded within commerce, or should it be managed through a dedicated, best-of-breed PIM?

When a dedicated PIM may make sense

Embedded PCM can remain a practical choice for organizations whose product content processes are closely aligned with their SAP Commerce environment. A dedicated PIM becomes worth considering when product information has a wider role across the enterprise.

Questions to consider include:

• Are multiple suppliers, business units, regions, or agencies contributing product information?
• Does the same catalog need to be adapted for different languages, markets, retailers, marketplaces, or distributors?
• Do marketing and product teams need greater ownership of enrichment, validation, approvals, and localization?
• Are custom workflows or data-quality rules becoming difficult to maintain within the commerce platform?
• Do AI initiatives require a more governed taxonomy, richer attributes, stronger context, and a reliable source of approved product information?
• Would separating product data from commerce make future platform changes faster and less disruptive?

These are stronger indicators of a PIM opportunity than the maintenance date alone. The goal is not to introduce another platform for its own sake. It is to create an operating model in which product information can be collected, enriched, governed, and activated efficiently wherever the business needs it.

Akeneo can complement whichever commerce path you choose

Akeneo provides a centralized source of trusted product information that can sit alongside SAP Commerce Cloud or support a broader commerce transformation. Product data can be brought together from ERP systems, suppliers, and internal teams; enriched and governed in one place; and then delivered to commerce platforms, marketplaces, distributors, and other customer-facing destinations.

For organizations continuing with SAP, Akeneo offers an accelerator for SAP Commerce Cloud built on SAP Integration Suite. It provides prebuilt integration artifacts to help synchronize complete, localized product information from Akeneo PIM into SAP Commerce Cloud while reducing the need for custom integration work.

This approach allows commerce and product information to evolve on their own timelines. Your commerce platform can focus on transactions, storefront experiences, promotions, and orders, while a dedicated PIM supports product data quality, enrichment workflows, governance, and omnichannel activation.

It also creates room for a phased modernization strategy. An organization may first migrate its commerce infrastructure, then strengthen product data governance, expand to new channels, improve supplier collaboration, or prepare its catalog for AI-powered experiences. Modernization does not have to be a single, disruptive replacement project.

Akeneo: a partner for life

Every SAP Commerce environment is different. Some organizations have relatively standard implementations; others have years of customizations, integrations, and business processes built around the platform. The right modernization plan must reflect that reality.

Akeneo is committed to being a partner for life, not simply a technology selected for a single migration project. We help organizations assess the current role of product information, define the capabilities their future architecture requires, and build a practical path forward, whether that means complementing SAP Commerce Cloud with a best-of-breed PIM or supporting a wider transformation.

The July 2026 SAP maintenance milestone deserves attention, but it does not need to create panic. Approached thoughtfully, it can become a catalyst for better governance, greater architectural flexibility, and more reliable product experiences. The result is a product information foundation that can support today’s migration decisions and continue adapting as channels, customer expectations, and AI capabilities evolve.

Considering your next step after SAP Commerce on-premise? Connect with an Akeneo expert for a conversation about your modernization strategy, or request a personalized demo to see how a best-of-breed PIM can support your business, whether you’re migrating to SAP Commerce Cloud or exploring a broader transformation.